1. Field of the Invention
The present invention relates to devices for converting between electrical and optical signals. More particularly, this invention relates to a method of manufacturing electro-optical devices.
2. Description of Related Art
Electro-optical modules can be used to convert electrical signals into optical signals and vice versa. Many different types of electro-optical modules are presently being manufactured. These modules have many applications, particularly within data-communications technology where electrical signals are carried by fiber optics, and can range in cost from under a hundred dollars to many thousands of dollars per module, depending on their application and functionality.
Different types of electro-optical modules can be used to perform different functions. Receive modules and transmit modules, for example, can each be used to provide half of an electro-optical conversion. More particularly, receive modules convert optical signals into electrical signals as part of a receive function. Transmit modules convert electrical signals into optical signals as part of a transmit function. Transceiver modules can be used to perform the electro-optical conversion for both receive and transmit paths. Transponder modules provide the same functionality as transceiver modules but also provide serialization and deserialization of the electrical signals.
Modules can be further categorized based on the type of light emitter used. Typical light emitters include surface emitting sources (such as Light Emitting Diodes (LEDs) and Vertical Cavity Surface Emitting Lasers (VCSELs)) and edge emitting sources (such as Fabry-Perot lasers and Distributed Feedback (DFB) lasers). Surface emitting light sources are generally used in the manufacture of low-cost modules.
The primary advantages of surface emitting light sources are in testing and assembly. Surface emitters can, for example, be easily tested on whole wafers. No assembly of the individual modules is therefore required before testing the part. The  assembly process is also simpler because the edges of the parts do not need to be polished. In addition, surface emitters can be easily assembled into arrays of multiple emitters (for example, a 1×12 array of VCSELs). Surface emitter arrays greatly simplify the assembly of parallel optical modules.
There are several challenges, however, in manufacturing optical modules. Among these challenges, it is difficult to align an optical fiber to an active optical area of a light emitter or a light detector. In addition, emitters may degrade or malfunction at relatively low temperatures, and removing heat from the emitters can be difficult. It is also difficult to test conventional modules. Another challenge is minimizing the number of parts required in the module assembly. A lack of “batch” manufacturing process steps also prevents lower cost manufacturing of modules. Some conventional solutions to these challenges are described briefly below.
Manufacturers of optical modules generally use some type of lens system (such as a spherical lens) to focus light into and out of the optical fiber. FIG. 1 is a schematic illustration of a conventional lens focusing system for an optical module. Referring to FIG. 1, this conventional focusing system includes a spherical lens 1, a multi-mode optical fiber 2 with an optical core 2a, and an electro-optical component 3 with active area 3a. These parts collectively form a system having an optical axis 4 and a ray trace 5. The surrounding material 6 is typically air.
The use of a lens is advantageous for at least two reasons. First, it acts as a light gathering element to collect the light from the emitting side. Second, it acts as a light focusing element to converge the light on the receiving side. These two actions result in a relaxation of the alignment tolerances between the emitting and receiving sides. In effect, the lens acts as a large target area for the emitter, when compared to the size of the receiver, while creating a focused spot that is small compared to the size of the receiver. Thus, the emitter may move around to some extent and still hit the lens, and the receiver may also move around to some extent and still have the focused spot fall within its active area.
Even with a lens, however, the alignment tolerances in a typical module require an active alignment process, which is conducted during the assembly of the optical system. In this process, the optical emitter is switched on and an output of the optical receivers is measured. The whole assembly is then micro-manipulated, typically by a human operator, to maximize the received signal by bringing all sub-components  into fine alignment. A flash cure process is then typically performed to freeze the assembly in place once fine alignment has been achieved.
There are disadvantages with the current methods for optical alignment. Optical elements such as glass lens arrays may be expensive. They are also typically small and may be delicate and difficult to handle and manipulate. Given that the optical assembly must be adjusted for fine alignment, some allowance must be made in the module design to facilitate this adjustment step. Active alignment is a slow, human driven process and is consequently expensive and error prone. Problems with optical alignment or failure to align correctly can create a significant rate of failure during the assembly process, further increasing the production costs of the module.
It would be advantageous to achieve optical alignment without an active alignment step through the inherent construction of the module (called “passive alignment”). It would be further advantageous if the optical components were very low cost and if there was no requirement to handle them as a separate sub-assembly.
Thermal management is also difficult in conventional optical modules. Optical transceiver or transponder modules, for example, typically require four different types of discrete semiconductor chips in close physical association with the optical axis. The light emitter and light detector are arranged on the optical axis. A driver chip for the emitter and an amplifier chip for the detector are also typically required. Emitters usually consume a significant amount of electrical power. The connection between the emitter and the driver chip is a major source of Electro-Magnetic Interference (EMI) and the signal degrades as the length of the connection increases. It is therefore advantageous to locate the emitter driver close to the emitter to limit the length of the connection. It is also advantageous to locate the receiver amplifier close to the detector because the detector output signal is very weak and therefore quickly degrades as the connection length increases.
These four semiconductor chips are therefore typically located in a very small area that is physically close to the emitter. Each chip consumes power that is dissipated as heat. In the case of a VCSEL emitter, for example, this heat may cause the VCSEL to function poorly in terms of its optical power output (slope efficiency), threshold current, and center wavelength accuracy. Or it may cause the VCSEL emitter to stop functioning completely. In addition, it may cause premature  aging and early failure of the VCSEL. It is therefore desirable to provide an efficient conduction path for drawing heat away from the VCSEL.
FIGS. 2A and 2B show typical thermal control solutions in a conventional optical module. More particularly, FIG. 2A illustrates four semiconductor devices 7 bonded with die attach material 8 to a circuit board 9. The circuit board 9 contains thermal vias 9a and is abutted to a metal heatsink 10. FIG. 2B illustrates two semiconductor components 11 bonded directly to a metal heatsink 13 with die attach material 12. In both of these examples, the metal heatsink is able to fairly efficiently transfer heat away from the semiconductor components to the cooler surrounding medium.
Conventional thermal management solutions, however, require that the semiconductor devices be bonded to thermally conductive materials. This may be a disadvantage if it prevents or impacts the ability to optically align or test the module. For example, bonding small optical components directly to a large heatsink requires manipulation of the entire heatsink in order to accurately align optical components.
In some cases, a separate, small circuit board may be used as an “optical substrate” on which optical die are mounted. The use of an optical substrate adds additional cost to the system, however, because it requires the formation of thermal vias through the substrate and possibly requires other thermal management. In that solution, a second circuit board, without additional thermal management features, is then required to route the signals from the optical substrate to the module output connector. This arrangement adds complexity, difficulty, and cost to the manufacturing process.
It would be advantageous if thermal management of the module could be accomplished without impacting the alignment of the module, while still providing the most optimal path for heat transfer away from the VCSEL. It would be further advantageous if a single circuit board could be used for the module to reduce complexity and cost.
Given the complexity and number of sub-components found in a typical module, some form of testing is important to guarantee a sufficient level of quality to the end user. Generally, every function of the module is tested. For example, the operation and performance parameters of the transmit function should be checked. These parameters include laser output power, extinction ratio, at-speed operation, and jitter performance. In many cases the drive parameters for each laser in the  module must be electrically adjusted to enable the module to meet its specifications over the full operating temperature range. The receiving (RX) channels are tested in a similar manner to verify performance to the required Bit Error Rate (BER). The module as an entire system should also be tested because the lenses and thermal performance also contribute to its overall performance. Module manufacturers therefore typically only test fully assembled modules. This test can take a long time (e.g., 10–20 minutes) to conduct and frequently requires a human operator to handle the module and conduct the test.
Currently, very few VCSEL suppliers are capable of guaranteeing zero defects for early life failures for unpackaged parts. Those that can make this guarantee charge a premium price for their product. Thus, module manufacturers who use VCSELs must burn-in their fully assembled modules for some period to eliminate modules whose VCSELs fail early in their life.
One of the fundamental disadvantages of conventional module designs is that you may only test and burn-in the fully assembled module. This means that defects that may occur at any point in the process are not detected until the final step. It further means then when a module fails, the full cost of the assembled module is lost. In addition, the modules are relatively large and difficult to handle with automated equipment and human intervention is therefore frequently required to insert and remove these modules from test fixtures.
It would be advantageous if the electro-optical components could be fully tested for alignment, operation, and performance as a single sub-assembly prior to final assembly into a finished module. Thus, the full cost of the module would not be incurred if the electro-optical sub-assembly failed. It would be further advantageous if the fully testable electro-optical sub-assembly were easy to handle and test with automated equipment. It would be advantageous if the electro-optical sub-assembly were compact and easy to store in standard sized component trays, which are used by most available semiconductor handling equipment. It would be advantageous if fully tested electro-optical sub-assemblies were able to be stored in inventory so that modules could be quickly assembled with various different styles of heatsink, which is a typical requirement of module customers.
There are also disadvantages with the current methods for manufacturing, as well as with the quality and reliability of the finished product. Conventional modules are generally manufactured manually. They also typically include multiple small  circuit boards and various other mechanical sub-components that must each be handled and placed in the module. There are no common form-factor methods (such as industry standard component trays that exist for semiconductors) for storing or handling the various sub-components of the optical modules. There is therefore little or no automated equipment available off the shelf to automate the assembly process. Because the process is primarily manual, it is prone to human error as well as poor process control. This impacts the manufacturing yield, as well as the quality and reliability of the product.
It would therefore be advantageous to have a module design that can be manufactured using a fully automated assembly process. It would be further advantageous if this assembly process could be performed using commonly available equipment, such as manufacturing equipment used in the semiconductor industry. It would also be advantageous if the assembly processes and materials used in the module assembly were commonly used and well understood. A module designed to have these advantages would be easier to manufacture at lower cost and at superior levels of yield, quality, and reliability.